A recent survey has shown that the EEG is of doubtful value as an outcome variable in clinical antiepileptic drug (AED) trials. Analysis of findings in 6 trials shows that in only two no consistent effect was seen; that in two power spectral analysis provided additional information to confirm changes in background activity; that in one a close relationship could be established between reduction in frequency of epileptiform discharges and administration of the trial drug (lamotrigine), and that in one, even though no correlation was apparent during use of the AED, there was an increase in frequency of both discharges and seizures on withdrawal (CGP 11952). In general the EEG would appear to be unsuitable as an outcome variable for assessing drug efficacy in AED trials. On the other hand it may give insight into the mode of drug action. Conceivably more efficient use could be made of the EEG if the methodology--including patient selection, consideration of circadian rhythms and of combination AED therapy, and standardized long-term recording--were to be improved and automation and quantification techniques used.